saveStateBanner Game Development

The Lost Gamestate…

Persisting state of things in an adventure is a really hard thing to crack. Is it coincidence that persisting and resisting sound the same? I waited some time to implement this because I needed to find out first how the bridging between the Objective C and the Lua code sorted out. Maybe I needed to store the state in anoyher way? And I didn’t had any experience with a dynamic language like Lua. But as I gained more experience with Lua I thought I was ready to implement the save games routines. But still… The choices you have to make are pretty hard because design mistakes like this will hunt you throughout the development of the game. I looked at various adventure projects currently in development and tried to find out how they had solved this or how they will solve this problem. Some of them solved this by serializing the state if the objects as they were loaded in memory. This is very commonly used in managed languages as Java and C#. In fact,every serious programming language has it’s own way of storing this. Objective C uses NSCoding to make the objects persist-able. I used this technique on earlier games I made but I hated the hard labour you had to put into it. You have to explicitly tell the system what to serialize and not to serialize. And with a complex hierarchy like adventure objects it can become a mess. So I tried to avoid this entirely. I started this project with Objective C for the graphics and heavy lifting. Lua (a great embedded language)  is used to write the behaviour of my game and manages all the state present in the adventure. My first thought was storing each Lua call like versioning systems do. Every time something happens the Lua call will be stored and when the game needs reloading all Lua calls will be executed in the order they were executed before. The big problem with this is that the entire history of calls are getting longer and it can hurt performance in the long run. The savegames in this scenario will mark several branching points. This could be a viable solution but besides the pros there are a lot of cons Pros: Only one big Lua State to manage Very accurate tracking of every possible state in the game Not much coding to do except loading and replaying Cons: Increasing loading and execution time before starting room Need to store the lua call immediately in memory or database Stores stuff that might not be necessary to be stored. Most of the solutions I encountered were storing the state of a subset of the Lua objects at specific intervals. This seemed a better approach to me as I had already chosen to make one Lua script per room. Lua fetches all object states when the room is entered and if it’s not present a template will be set. Right now I can differentiate between 4 kinds of state objects: RoomState : Stores all the stuff that happens in the room the player is in. PlayerState : The player (Playable Character) moves between rooms and will be loaded beforehand (to see which room he is and will be his starting point). NpcState : Some Npc’s (Non Playable Characters) will and can move across different rooms, and they will be added and loaded according their currentroom property. GlobalState : All event data that needs to be accessible at any time. This will for example be events setting a state variabele that are triggered while having a dialog with a npc and some other dialog depends on that state The next part of the solution is in which format I can store the lua state. That decision was easy: Json is superefficient because the way lua tables are initialized is json without quotes. I found a very basic json implementation that worked without much fuss. Look at the difference between Lua table creation and it’s Json format :

It’s basically replacing the equal sign for an semicolon sign and wrap the props in quotes and you have the Json format basically. I used this Basic Json parser that parses json to tables and tables to json. The only problem I had was that the parser tried to serialize the functions too.  This could lead to crashes because I could not load the function back from the database. I wondered if I could seperate the data from the functions. And yes, this can be done really nicely with the use of Lua metatables. A metatable can be seen as an extension point when set on a table. A very good tutorial on how this works can be found here. The structure I came up looks like this:

The showed function is the ability for the object to persist itself.  The luaController is a user data object that calls the Objective C method to persist the Json string. This defined metatable with the persist function can be set on the player at any time. That way I could create some ObjectOriented approach to handle these kind of things.

The last thing I needed to figure out was how to store this. I chose SSLite because this is a well supported choice on different platforms (android needs to be supported too). I made a nice wrapper for crud operations. When a save-game is created by the player the intermediate and temporary json records are copied to a different table. It will get an additional timestamp, title and a save image that consist of a freeze frame screenshot. Well this is the plan, that part is not implemented yet. But I have to start somewhere don’t I?

luaBanner Game Development

Lua Scripting

I started off early with Lua scripting. This was mainly because the Ray Wenderlich RPG samples were using it and this gave me a head start and a direction. I think Lua is a beautiful scripting language, and it has real potential (see the Love2D framework). It’s what I can teach my 9 year old son. So it’s pretty nice. I am still in pre-production and I wanted to make coding the actor (NPC and player) actions as easy as possible. If you settle to early for something simple to progress quickly, it can be really cumbersome when you need to make a lot of changes to rooms and actors. I made a pickup animation sequence completely in Objective C, but I knew this was wrong. If I had to change or extends this I had to change about 40 lines of code and it wasn’t flexibal. This was wrong because the behavior of actors is data but in the form of scripts. I have one script per room. As you can see this is much nicer and flexible :

This script will : Walk the main character to the item. He will kneel to pick it up and will rise again. This is a back and forth animation Add the item to the inventory of the player Wait for about a second and run a lua method called passHammerToDave that alters the state of the character. Dave will ‘own’ the item now. Now it passes all these actions to the CCActionSequence that is supported by the Cocos2D framework. The seqOpen will be passed as an array of commands to the Objective C bridging code, and this will execute this in sequence as CCAction’s! Many actions in cocos2d are finite time actions, that means that walking to a certain tile with the help of an A* (Astar) algorithm does not fit easily into an action. So I splitted all separate CCAction sequences and let every sequence end with an action to fire an event to a sort of action manager class that starts the next sequence. This looks like this :

A lot of code, for sure. But this is what I need to do once… Maybe the code needs some cleaning up, but for now it’s ok. It does the job well. Every action will be perfectly in sequence because each action will fire an event to the characterAnimationManager to signal that the next sequence should be executed. The next scripts will show the scripting for opening the elevator door and going to another room :

The script does the following steps: The player walks to the button location that is defined in the room he is in. The player will reach for the button when it uses the verb “use” on the control panel Button sound plays (this is alway an asynchronous event) The sound for sliding the elevator is played The elevator has some tile locations that needs to be blocked or unblocked when doors close or open up. The sequence runs the open elevator door animations that is called via lua via function ‘openElevatorDoor’ Well lets see these both scripts in action! (trust me… it’s quite short) : Scripting from franzzle on Vimeo.

Screenshot Sunday Game Development

App Icon Evolution

Is it too early to look at an icon? I think you need to start early, trying different designs and get into grips with a certain look… It’s these kind of things that, if you wait too long for it to make and when you are trying to finish the game, you will not make a proper icon. Those final crunch days are gruesome because bugs need to be squashed, menu’s have to be made, a settings screen you forgot and maybe the gameplay still lacks some finesse? Eventually  you will not have the time to make it anyway! One other thing is that you need to know how long your icon could last. Does it bore you? Does it grow on you? Does it stand out? I am on to my third icon now, and it could be the final one. The icon below was a really early icon I created just to try to lay out the composition and to get off the default Spritebuilder icon, Dave was standing against the wall ready to be fusilladed by german WWII soldiers. I you think this looks familiar, it is! Because a part of the cartoon is actually stolen from the political cartoon for the Peking Olympics The idea was that Dave would transported back to the present day with a kind of transporter beam. I knew this was only a working icon and I would make a complete overhaul of it. But the transporter thingy, stuck with me though! A long period after that I thought that the Time Machine had a shape that could be fitted into an icon. So I made a Hires character and put that in the time machine. This was a first draft and the dave character was not recognizable on small icons. The colors were completely off and it did not have the pixelated look I was looking for. So I came up with this icon. I still use it today. The character is the Dave character as he will walk around in the game. The colours are bright and playful. The transporter twinkling stars makes it as if Dave is actually being transported back in time! Maybe I should tweak the shadowing a bit? Well, that’s it for now!  

Created a build number script to The Lost Adventures title screen

I was sick of guessing how recent the installed version of The Lost Adventures was on my iDevices. So I made an Xcode pre-build script that writes the Git commit number plus a date/time  to a simple txt file. This is picked up and shown on the main title screen. I thought I did more commits last year… But maybe I didn’t commit at the start of the project…? That could be… Vacation….? Anyway, nice…!  

Screenshot Sunday Game Development

Screenshot Sunday : When you gotta go, you gotta go

One of the rooms in The Lost Adventures will be the mens toilets. If I tell What Dave needs to do here I would  spoil one of the puzzles of the game. No spoilers here !. You can obviously try to “use” the toilet, but supporting this action would take me hours of animating those 9 frames to sit down and what to think about those frames for pulling the chord to flush? But who knows? When I have some more common animations for Dave, maybe…. Oh and do you see that white crosshair? I was watching the GUI of the Fate of Atlantis and this was still missing in my adventure game. The default action will be the “Walk to” action (not a button on the screen, but still an action). When you hover over the items, the label of a POI (Point of Interest) will be shown. This presented a problem on the iPad though… Your hand blocks the visibility of the label, so I need to place the label to the left or right. Or I will place the label-bar to the top especially for the iPad. Enough tasks left to improve though. I am still in the preproduction stage (and this is actually good!!). Polishing a game and making content  is much harder and more tedious! Luckily all rooms are in draft now, to support the entire story for Level 1. When the work on the adventure engine is finished and I have made one playable level, I will be in production stage. One of the main task is passing the list of actions to the Lua scripting engine. I have greatly improved the way the scripting works. The drawback of an earlier approach meant timing the various actions exactly before the next scripted action started. I needed to solve this because the time taken by walking through the room was not predictable. Now the character can walk to a certain tile, then proceed to a new action like picking up an item (this is calculated by number of frames * frame-delay). These walking and animation actions are handled completely in sequence. Playing a sound can be a simultaneous action. The next blog post will be entirely about this nifty system.

Recent Posts

Recent Comments